Within the field of remote sensing, Unmanned Aerial Vehicle (UAV) infrared object detection plays a pivotal role, especially in complex environments. However, existing methods face challenges such as insufficient accuracy or low computational efficiency, particularly in the detection of small objects. This paper proposes a lightweight and accurate UAV infrared object detection model, YOLO-UIR, for small object detection from a UAV perspective. The model is based on the YOLO architecture and mainly includes the Efficient C2f module, lightweight spatial perception (LSP) module, and bidirectional feature interaction fusion (BFIF) module. The Efficient C2f module significantly enhances feature extraction capabilities by combining local and global features through an Adaptive Dual-Stream Attention Mechanism. Compared with the existing C2f module, the introduction of Partial Convolution reduces the model’s parameter count while maintaining high detection accuracy. The BFIF module further enhances feature fusion effects through cross-level semantic interaction, thereby improving the model’s ability to fuse contextual features. Moreover, the LSP module efficiently combines features from different distances using Large Receptive Field Convolution Layers, significantly enhancing the model’s long-range information capture capability. Additionally, the use of Reparameterized Convolution and Depthwise Separable Convolution ensures the model’s lightweight nature, making it highly suitable for real-time applications. On the DroneVehicle and HIT-UAV datasets, YOLO-UIR achieves superior detection performance compared to existing methods, with an mAP of 71.1% and 90.7%, respectively. The model also demonstrates significant advantages in terms of computational efficiency and parameter count. Ablation experiments verify the effectiveness of each optimization module.

Surface defect detection is an important task in industrial production. Although significant progress has been made in deep learning-based defect segmentation methods, the low contrast between defect and background and the shape and scale diversity of defects limit the models’ detection accuracy and generalization ability. Therefore, realizing the full automation of surface defect detection still faces many challenges. To overcome these problems, this paper proposes a skeleton-strengthening network called SKS-Net, which provides stable and excellent surface defect detection performance even at low contrast and multi-scale. SKS-Net designs a skeleton-strengthening convolutional module to capture multi-scale features efficiently. The convolution kernel of this convolution module is closer to the shape of the segmentation target, which significantly reduces irrelevant regions and improves the feature extraction capability of the convolution kernel. This convolution module can be directly embedded into existing network structures without adding additional computational overhead. In addition, we design a new feature channel fusion module to extract key information from features at different levels. To improve the training effect, we introduce a multi-scale auxiliary supervision mechanism. The proposed model is evaluated on four different publicly available surface defect datasets and compared with other state-of-the-art models. Results show that SKS-Net performs exceptionally well in terms of accuracy, achieving 66.72% mIoU on the KolektorSDD dataset. The code is publicly available at https://github.com/Wanglaoban3/SKS-Net.git.

Weld defect detection is an important task in the welding process. Although there are many excellent weld defect detection models, there is still much room for improvement in stability and accuracy. In this study, a lightweight deep learning model called WeldNet is proposed to improve the existing weld defect recognition network for its poor generalization performance, overfitting and large memory occupation, using a design with a small number of parameters but with better performance. We also proposed ensemble-distillation strategy in the training process, which effectively improved the accuracy rate and proposed an improved model ensemble scheme. The experimental results show that the final designed WeldNet model performs well in detecting weld defects and achieves state-of-the-art performance. Its number of parameters is only 26.8\% of that of ResNet18, but the accuracy is 8.9\% higher, while achieving 41 FPS on cpu to meet the demand of real-time operation. The study is of guiding significance for solving practical problems in weld defect detection, and provides new ideas for the application of deep learning in industry. The code used in this article is available at: https://github.com/Wanglaoban3/WeldNet.git

Quality inspection of industrial products is a critical task to ensure that products meet the required standards. Although deep learning-based detection models have achieved significant progress in recent years, the limited availability of samples in real industrial environments still constrains the generalizability of these models. In this paper, we propose a novel semi-supervised defect segmentation framework that leverages contrastive learning to handle a large number of unlabeled samples. The core of this framework is the contrastive consistency regularization loss and class affinity loss, which together enhance the model’s ability to learn from unlabeled data. Our proposed contrastive consistency regularization loss incorporates pixel-level features from the entire batch into a cross-image consistency objective, substantially enhancing the model’s robustness to appearance variations. Complementarily, the class affinity loss leverages a compact memory bank of high-quality prototypes to directly refine the feature space—pulling unlabeled pixels toward their probable class centroids and pushing them away from alternative classes—thereby tightening intra-class compactness and sharpening inter-class boundaries for superior segmentation accuracy. The experimental results indicate that in the three different available labeled data experiments on NEU-SEG dataset, our proposed method improves the mIoU of the supervised learning paradigm by 4.4%, 3.7%, and 3.8%, respectively, surpassing the previous state-of-the-art method by 0.6%, 0.7%, and 0.5%. Finally, the ablation studies validate the effectiveness of the proposed method.

In order to improve the accuracy of industrial defect detection based on deep learning, the special defect data set IDEF (Industrial defect data set, IDEF) is constructed, and the defect data is enhanced by digital image processing and deep learning. After data enhancement, the accuracy of defect detection is improved by 0.55 percentage points. Aiming at the phenomenon that small target defects are easy to be missed by the model, a method of adding grid intersection points to the regression strategy of target detection network is proposed to increase the predicted defect benchmark. The experiment shows that this method improves the defect detection accuracy map (mean average precision) by 1.96%. Aiming at the problem of poor representation ability of small target defects, a scale weight loss function is proposed, which makes the model more sensitive to defects with smaller size and more thorough learning of small target defects. It is found that this method improves the accuracy of defect detection by 2.23%. Finally, based on the fcosv2 basic model, two small target defect detection strategies are used at the same time to improve the defect detection accuracy by 3.35%.